Skip to content
RobotWorld
← Back to Videos

Agentic Real2Sim: reconstructing a physics simulator from a single video

Loading video
Loading video

Agentic Real2Sim lets a vision-language agent reconstruct an editable, executable physics simulator from a single real-world video, instead of learning physics directly from video.

World ModelsAgentic Real2SimPhysics simulationVLAResearch
Category: research
Author: @robotsdigest
Date: 2026-07-22T00:00:00
Duration: 254.72s
Source: https://x.com/robotsdigest/status/2079981083622576623

Related Videos

Why Reward-Based RL Is Extremely Sample Inefficient
0:06

Why Reward-Based RL Is Extremely Sample Inefficient

@cybernetic_lab · 459 views · 2026-07-31
Reinforcement LearningSample EfficiencyWorld Models
FLUX 3: Black Forest Labs launches real world model
0:46

FLUX 3: Black Forest Labs launches real world model

@shao__meng · 590 views · 2026-07-24
FLUX 3Black Forest LabsWorld Models
Astribot releases Lumo-2 World-Action Model
0:39

Astribot releases Lumo-2 World-Action Model

@LeoKharon · 492 views · 2026-07-23
WAMLumo-2Astribot
AlayaWorld: Long-horizon playable video world generation (15B)
0:19

AlayaWorld: Long-horizon playable video world generation (15B)

@kwangmoo_yi · 466 views · 2026-07-22
World ModelsAlayaWorldVideo Generation
ABot-World-0: Infinite interactive world rollout on a single desktop GPU
0:05

ABot-World-0: Infinite interactive world rollout on a single desktop GPU

@_akhaliq · 453 views · 2026-07-22
World ModelsABot-World-0Video Generation
Patch Policy: pretrained ViT beats OpenVLA-OFT with 0.7% params
0:22

Patch Policy: pretrained ViT beats OpenVLA-OFT with 0.7% params

@jeffacce · 391 views · 2026-07-22
VLAPatch PolicyViT

As an Amazon Associate, we earn from qualifying purchases.